gemma-4-12B-it-qat-w4a16-ct-heretic API & Inference Endpoint

Load it the same way

python
from transformers import AutoProcessor, Gemma4UnifiedForConditionalGeneration

repo_id = "wnfldchen/gemma-4-12B-it-qat-w4a16-ct-heretic"

processor = AutoProcessor.from_pretrained(repo_id)
model = Gemma4UnifiedForConditionalGeneration.from_pretrained(
    repo_id,
    device_map="auto",
)

What is included

Compressed model weights in standard Hugging Face transformers layout.
config.json, tokenizer files, processor_config.json, and chat_template.jinja.
The exact compression recipe as recipe.yaml.

Provenance

Source uncompressed Heretic model: wnfldchen/gemma-4-12B-it-qat-q4_0-unquantized-heretic
Quantization: 4-bit int grouped weights, W4A16-style compressed tensors
Group size: 32
Saved format: pack-quantized
Transformers model type: gemma4_unified
Compression recipe preserved as recipe.yaml

W4A16 reproduction

[!TIP] This model is reproducible.

Recreate the public unquantized Heretic source checkpoint pinned below, run the recorded compression command, and verify the resulting artifacts against the hashes in this section.

Source checkpoint

Source unquantized Heretic model: wnfldchen/gemma-4-12B-it-qat-q4_0-unquantized-heretic
Source repo revision pinned at upload time: ca3b6a31f6566df945912929a5964e6d91e918c9
Source model card: README
Source reproducibility guide: README
Compression recipe preserved in this repo as recipe.yaml

Compression command

This is the command to recreate this compressed export from the unquantized Heretic source model:

bash
python scripts/compress_to_w4a16.py \
  wnfldchen/gemma-4-12B-it-qat-q4_0-unquantized-heretic \
  --output ./gemma-4-12B-it-qat-w4a16-ct-heretic \
  --device-map cuda:0

The helper defaults --dtype to auto; the recipe itself is preserved as recipe.yaml.

Compression environment

Platform: Linux-6.8.0-111-generic-x86_64-with-glibc2.39

Table with columns: Component, Version
Component	Version
`Python`	`3.12.13`
`torch`	`2.12.0+cu130`
`transformers`	`5.10.1`
`llmcompressor`	`0.12.0`
`compressed-tensors`

Verification

Recreate the export, then compare the generated files with the SHA256 hashes below.

README.md is intentionally omitted because this model card contains the reproducibility data and therefore cannot hash itself deterministically.

Table with columns: File, SHA256
File	SHA256
`chat_template.jinja`	`ae53464bf3be25802b3a5b37def7fd89667067d7577049b3b2d74c4d8de4c6d4`
`config.json`	`90cd1ae2fdab0741f1fbbb47b399e8ec19b6a657c999fd41450dedf1933170ba`
`generation_config.json`	`0f809910f55b332e334eebf13dc415730e316d187c38d1367dc5b4cd5e27fc25`
`model.safetensors`	`261576a4588f5d7d7e0d220f8761cb7bcd71e815f733fc8c3ed3da3d1c2b79cd`

Source model card reference

The section below is copied from the source unquantized Heretic export for reference. That source card already embeds the upstream Google Gemma 4 model card beneath its own note.

Unless it explicitly says otherwise, benchmark tables, hashes, and abliteration details there refer to the pre-compression checkpoint, not this W4A16 repack.

This is a decensored version of google/gemma-4-12B-it-qat-q4_0-unquantized, made using Heretic v1.4.0 with the Arbitrary-Rank Ablation (ARA) method (with row-norm preservation)

[!TIP] This model is reproducible!

See the README in the reproduce directory for more information.

Abliteration parameters

Table with columns: Parameter, Value
Parameter	Value
start_layer_index	19
end_layer_index	42
preserve_good_behavior_weight	0.9174
steer_bad_behavior_weight	0.0004
overcorrect_relative_weight	0.9004
neighbor_count	13

Performance

Table with columns: Metric, This model, Original model (google/gemma-4-12B-it-qat-q4_0-unquantized)
Metric	This model	Original model (google/gemma-4-12B-it-qat-q4_0-unquantized)
Keywords	11/100	99/100
KL divergence	0.0606	0 (by definition)

Evaluation

The edited model was compared against the original base model using Heretic's integrated lm_eval benchmark flow. Results below are reported exactly from the benchmark run, with an added delta column for quick comparison.

Delta = edited model minus original model. Negative values mean the edited model scored lower.

Table with columns: Benchmark, Samples, Metric, This model, Original model, Delta, This stderr, Original stderr
Benchmark	Samples	Metric	This model	Original model	Delta	This stderr	Original stderr
WinoGrande	1267	Accuracy	49.49%	50.51%	-1.02 pp	1.41 pp	1.41 pp
PIQA	1838	Accuracy	53.48%

Interpretation

Across these benchmarks, the edited model scores lower on four reported metrics and higher on two relative to the original model.

The largest decline is -1.23 pp on CommonsenseQA accuracy. The largest improvement is +9.94 pp on EQ-Bench parseable outputs.

These results should still be interpreted cautiously. The benchmark set here is small, and the metrics are not all on the same scale, so changes are best read within each benchmark rather than as a single aggregate score.

Notes

WinoGrande, PIQA, and CommonsenseQA are reported as accuracy-style metrics.
PIQA includes both standard accuracy and normalized accuracy, following lm_eval output.
EQ-Bench is reported with both the benchmark's main score and the percentage of parseable outputs.
Percentage-style standard errors are shown in percentage points.
These results reflect this specific benchmark run and evaluation configuration; broader capability conclusions should ideally be supported with additional benchmarks.

[!NOTE] The remainder of this README reproduces the upstream model card for the base checkpoint. Any benchmark tables or capability claims below this point refer to the upstream base model, not this Heretic-edited derivative, unless explicitly stated.

[!Note] This model card is for the new versions of the Gemma 4 family optimized with Quantization-Aware Training (QAT), which allows preserving similar quality to bfloat16 while dramatically reducing the memory requirements to load the model. Four versions of the QAT checkpoints are available:

Unquantized QAT checkpoints (Q4_0): Half-precision weights extracted from the QAT pipeline, ideal for custom downstream compilation and research. Available for Gemma 4 E2B, E4B, 12B, 26B A4B, and 31B, and their drafter models.

GGUF (Q4_0): Ready-to-deploy formats for broad ecosystem compatibility. Available for Gemma 4 E2B, E4B, 12B, 26B A4B, and 31B.

Mobile-optimized (wNa8o8): A custom schema engineered explicitly for mobile hardware efficiency. It features targeted 2-bit decoding layers, optimized KV caches, and static activations to maximize VRAM savings. Available for Gemma 4 E2B and E4B.

Compressed Tensors (w4a16): QAT checkpoints serialized in the compressed-tensors format for native, optimized inference with vLLM. Available for Gemma 4 E2B, E4B, 12B, and 31B.

Assistant Compatibility: When using multi-token prediction (speculative decoding) with an assistant model alongside a QAT target model, the assistant model must also be a QAT checkpoint with the same precision to ensure compatibility.

Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages.

Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in five distinct sizes: E2B, E4B, 12B, 26B A4B, and 31B. Their diverse sizes make them deployable in environments ranging from high-end phones to laptops and servers, democratizing access to state-of-the-art AI.

Gemma 4 introduces key capability and architectural advancements:

Reasoning – All models in the family are designed as highly capable reasoners, with configurable thinking modes.
Extended Multimodalities – Processes Text, Image with variable aspect ratio and resolution support (all models), Video, and Audio (featured natively on the E2B, E4B, and 12B models).
Diverse & Efficient Architectures – Offers Dense and Mixture-of-Experts (MoE) variants of different sizes for scalable deployment.
Optimized for On-Device – Smaller models are specifically designed for efficient local execution on laptops and mobile devices.
Increased Context Window – The small models feature a 128K context window, while the medium models support 256K.
Enhanced Coding & Agentic Capabilities – Achieves notable improvements in coding benchmarks alongside native function-calling support, powering highly capable autonomous agents.
Native System Prompt Support – Gemma 4 introduces native support for the system role, enabling more structured and controllable conversations.

Models Overview

Gemma 4 models are designed to deliver frontier-level performance at each size, targeting deployment scenarios from mobile and edge devices (E2B, E4B) to consumer GPUs and workstations (12B, 26B A4B, 31B). They are well-suited for reasoning, agentic workflows, coding, and multimodal understanding.

The models employ a hybrid attention mechanism that interleaves local sliding window attention with full global attention, ensuring the final layer is always global. This hybrid design delivers the processing speed and low memory footprint of a lightweight model without sacrificing the deep awareness required for complex, long-context tasks. To optimize memory for long contexts, global layers feature unified Keys and Values, and apply Proportional RoPE (p-RoPE).

Dense Models

Table with columns: Property, E2B, E4B, 12B Unified, 31B Dense
Property	E2B	E4B	12B Unified	31B Dense
Total Parameters	2.3B effective (5.1B with embeddings)	4.5B effective (8B with embeddings)	11.95B	30.7B
Layers	35	42	48	60
Sliding Window	512 tokens	512 tokens	1024 tokens	1024 tokens

The "E" in E2B and E4B stands for "effective" parameters. The smaller models incorporate Per-Layer Embeddings (PLE) to maximize parameter efficiency in on-device deployments. Rather than adding more layers or parameters to the model, PLE gives each decoder layer its own small embedding for every token. These embedding tables are large but are only used for quick lookups, which is why the effective parameter count is much smaller than the total.

The "Unified" in Gemma 4 12B Unified refers to its encoder-free architecture. Other Gemma 4 models use dedicated encoders to process multimodal data before passing it to the LLM. Gemma 4 12B eliminates these encoders entirely, projecting raw image patches and audio waveforms directly into the LLM's embedding space through lightweight linear layers. This unified approach means all modalities flow straight into a single decoder-only transformer, reducing multimodal latency and allowing the entire model to be fine-tuned in one pass.

Mixture-of-Experts (MoE) Model

Table with columns: Property, 26B A4B MoE
Property	26B A4B MoE
Total Parameters	25.2B
Active Parameters	3.8B
Layers	30
Sliding Window	1024 tokens
Context Length	256K tokens
Vocabulary Size	262K
Expert Count	8 active / 128 total and 1 shared
Supported Modalities	Text, Image

The "A" in 26B A4B stands for "active parameters" in contrast to the total number of parameters the model contains. By only activating a 4B subset of parameters during inference, the Mixture-of-Experts model runs much faster than its 26B total might suggest. This makes it an excellent choice for fast inference compared to the dense 31B model since it runs almost as fast as a 4B-parameter model.

Benchmark Results

These models were evaluated against a large collection of different datasets and metrics to cover different aspects of text generation. Evaluation results marked in the table are for instruction-tuned models.

Table with columns: Gemma 4 31B, Gemma 4 26B A4B, Gemma 4 12B Unified, Gemma 4 E4B, Gemma 4 E2B, Gemma 3 27B (no think)
	Gemma 4 31B	Gemma 4 26B A4B	Gemma 4 12B Unified	Gemma 4 E4B	Gemma 4 E2B	Gemma 3 27B (no think)
MMLU Pro	85.2%	82.6%	77.2%	69.4%	60.0%	67.6%
AIME 2026 no tools	89.2%	88.3%	77.5%	42.5%	37.5%

*Excluding Chinese language.

Core Capabilities

Gemma 4 models handle a broad range of tasks across text, vision, and audio. Key capabilities include:

Thinking – Built-in reasoning mode that lets the model think step-by-step before answering.
Long Context – Context windows of up to 128K tokens (E2B/E4B) and 256K tokens (12B, 26B A4B/31B).
Image Understanding – Object detection, Document/PDF parsing, screen and UI understanding, chart comprehension, OCR (including multilingual), handwriting recognition, and pointing. Images can be processed at variable aspect ratios and resolutions.
Video Understanding – Analyze video by processing sequences of frames.
Interleaved Multimodal Input – Freely mix text and images in any order within a single prompt.
Function Calling – Native support for structured tool use, enabling agentic workflows.
Coding – Code generation, completion, and correction.
Multilingual – Out-of-the-box support for 35+ languages, pre-trained on 140+ languages.
Audio (E2B, E4B, and 12B only) – Automatic speech recognition (ASR) and speech-to-translated-text translation across multiple languages.

Getting Started

You can use all Gemma 4 models with the latest version of Transformers. To get started, install the necessary dependencies in your environment:

pip install -U transformers torch accelerate

Once you have everything installed, you can proceed to load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID,
    dtype="auto",
    device_map="auto"
)

Once the model is loaded, you can start generating output:

python
# Prompt
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Write a short joke about saving RAM."},
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
    enable_thinking=False
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=1024)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

To enable reasoning, set enable_thinking=True and the parse_response function will take care of parsing the thinking output.

Below, you will also find snippets for processing audio (E2B, E4B, 12B only), images, and video alongside text:

Make sure to install the following packages:

pip install -U transformers torch torchvision librosa accelerate

You can then load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, 
    dtype="auto", 
    device_map="auto"
)

Once the model is loaded, you can start generating output by directly referencing the audio URL in the prompt:

python
# Prompt - add audio after text
messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\n* Only output the transcription, with no newlines.\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three."},
            {"type": "audio", "audio": "https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/journal1.wav"},
        ]
    }
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=512)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

Make sure to install the following packages:

pip install -U transformers torch torchvision accelerate

You can then load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, 
    dtype="auto", 
    device_map="auto"
)

Once the model is loaded, you can start generating output by directly referencing the image URL in the prompt:

python
# Prompt - add image before text
messages = [
    {
        "role": "user", "content": [
            {"type": "image", "url": "https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/GoldenGate.png"},
            {"type": "text", "text": "What is shown in this image?"}
        ]
    }
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=512)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

Make sure to install the following packages:

pip install -U transformers torch torchvision librosa accelerate

You can then load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, 
    dtype="auto", 
    device_map="auto"
)

Once the model is loaded, you can start generating output by directly referencing the video URL in the prompt:

python
# Prompt - add video before text
messages = [
    {
        'role': 'user',
        'content': [
            {"type": "video", "video": "https://github.com/bebechien/gemma/raw/refs/heads/main/videos/ForBiggerBlazes.mp4"},
            {'type': 'text', 'text': 'Describe this video.'}
        ]
    }
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=512)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

Best Practices

For the best performance, use these configurations and best practices:

1. Sampling Parameters

Use the following standardized sampling configuration across all use cases:

temperature=1.0
top_p=0.95
top_k=64

2. Thinking Mode Configuration

Compared to Gemma 3, the models use standard system, assistant, and user roles. To properly manage the thinking process, use the following control tokens:

Trigger Thinking: Thinking is enabled by including the <|think|> token at the start of the system prompt. To disable thinking, remove the token.
Standard Generation: When thinking is enabled, the model will output its internal reasoning followed by the final answer using this structure:
<|channel>thought\n[Internal reasoning]<channel|>
Disabled Thinking Behavior: For all models except for the E2B and E4B variants, if thinking is disabled, the model will still generate the tags but with an empty thought block:
<|channel>thought\n<channel|>[Final answer]

[!Note] Note that many libraries like Transformers and llama.cpp handle the complexities of the chat template for you.

3. Multi-Turn Conversations

No Thinking Content in History: In multi-turn conversations, the historical model output should only include the final response. Thoughts from previous model turns must not be added before the next user turn begins, with the exception of tool call turns where thinking content should be preserved.

4. Modality order

For optimal performance with multimodal inputs, place:

Image content before the text in your prompt.
Audio content after the text in your prompt.

5. Variable Image Resolution

Aside from variable aspect ratios, Gemma 4 supports variable image resolution through a configurable visual token budget, which controls how many tokens are used to represent an image. A higher token budget preserves more visual detail at the cost of additional compute, while a lower budget enables faster inference for tasks that don't require fine-grained understanding.

The supported token budgets are: 70, 140, 280, 560, and 1120.
- Use lower budgets for classification, captioning, or video understanding, where faster inference and processing many frames outweigh fine-grained detail.
- Use higher budgets for tasks like OCR, document parsing, or reading small text.

6. Audio

Use the following prompt structures for audio processing:

Audio Speech Recognition (ASR)

text
Transcribe the following speech segment in {LANGUAGE} into {LANGUAGE} text.

Follow these specific instructions for formatting the answer:
* Only output the transcription, with no newlines.
* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three.

Automatic Speech Translation (AST)

text
Transcribe the following speech segment in {SOURCE_LANGUAGE}, then translate it into {TARGET_LANGUAGE}.
When formatting the answer, first output the transcription in {SOURCE_LANGUAGE}, then one newline, then output the string '{TARGET_LANGUAGE}: ', then the translation in {TARGET_LANGUAGE}.

7. Audio and Video Length

All models support image inputs and can process videos as frames whereas the E2B, E4B, and 12B models also support audio inputs. Audio supports a maximum length of 30 seconds. Video supports a maximum of 60 seconds assuming the images are processed at one frame per second.

Model Data

Data used for model training and how the data was processed.

Training Dataset

Our pre-training dataset is a large-scale, diverse collection of data encompassing a wide range of domains and modalities, which includes web documents, code, images, audio, with a cutoff date of January 2025. Here are the key components:

Web Documents: A diverse collection of web text ensures the model is exposed to a broad range of linguistic styles, topics, and vocabulary. The training dataset includes content in over 140 languages.
Code: Exposing the model to code helps it to learn the syntax and patterns of programming languages, which improves its ability to generate code and understand code-related questions.
Mathematics: Training on mathematical text helps the model learn logical reasoning, symbolic representation, and to address mathematical queries.
Images: A wide range of images enables the model to perform image analysis and visual data extraction tasks.

The combination of these diverse data sources is crucial for training a powerful multimodal model that can handle a wide variety of different tasks and data formats.

Data Preprocessing

Here are the key data cleaning and filtering methods applied to the training data:

CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was applied at multiple stages in the data preparation process to ensure the exclusion of harmful and illegal content.
Sensitive Data Filtering: As part of making Gemma pre-trained models safe and reliable, automated techniques were used to filter out certain personal information and other sensitive data from training sets.
Additional methods: Filtering based on content quality and safety in line with our policies.

Ethics and Safety

As open models become central to enterprise infrastructure, provenance and security are paramount. Developed by Google DeepMind, Gemma 4 undergoes the same rigorous safety evaluations as our proprietary Gemini models.

Evaluation Approach

Gemma 4 models were developed in partnership with internal safety and responsible AI teams. A range of automated as well as human evaluations were conducted to help improve model safety. These evaluations align with Google’s AI principles, as well as safety policies, which aim to prevent our generative AI models from generating harmful content, including:

Content related to child sexual abuse material and exploitation
Dangerous content (e.g., promoting suicide, or instructing in activities that could cause real-world harm)
Sexually explicit content
Hate speech (e.g., dehumanizing members of protected groups)
Harassment (e.g., encouraging violence against people)

Evaluation Results

For all areas of safety testing, we saw major improvements in all categories of content safety relative to previous Gemma models. Overall, Gemma 4 models significantly outperform Gemma 3 and 3n models in improving safety, while keeping unjustified refusals low. All testing was conducted without safety filters to evaluate the model capabilities and behaviors. For both text-to-text and image-to-text, and across all model sizes, the model produced minimal policy violations, and showed significant improvements over previous Gemma models' performance.

Usage and Limitations

These models have certain limitations that users should be aware of.

Intended Usage

Multimodal models (capable of processing vision, language, and/or audio) have a wide range of applications across various industries and domains. The following list of potential uses is not comprehensive. The purpose of this list is to provide contextual information about the possible use-cases that the model creators considered as part of model training and development.

Content Creation and Communication
- Text Generation: These models can be used to generate creative text formats such as poems, scripts, code, marketing copy, and email drafts.
- Chatbots and Conversational AI: Power conversational interfaces for customer service, virtual assistants, or interactive applications.
- Text Summarization: Generate concise summaries of a text corpus, research papers, or reports.
- Image Data Extraction: These models can be used to extract, interpret, and summarize visual data for text communications.
- Audio Processing and Interaction: The E2B, E4B, and 12B models can analyze and interpret audio inputs, enabling voice-driven interactions and transcriptions.
Research and Education
- Natural Language Processing (NLP) and VLM Research: These models can serve as a foundation for researchers to experiment with VLM and NLP techniques, develop algorithms, and contribute to the advancement of the field.
- Language Learning Tools: Support interactive language learning experiences, aiding in grammar correction or providing writing practice.
- : Assist researchers in exploring large bodies of text by generating summaries or answering questions about specific topics.

Limitations

Training Data
- The quality and diversity of the training data significantly influence the model's capabilities. Biases or gaps in the training data can lead to limitations in the model's responses.
- The scope of the training dataset determines the subject areas the model can handle effectively.
Context and Task Complexity
- Models perform well on tasks that can be framed with clear prompts and instructions. Open-ended or highly complex tasks might be challenging.
- A model's performance can be influenced by the amount of context provided (longer context generally leads to better outputs, up to a certain point).
Language Ambiguity and Nuance
- Natural language is inherently complex. Models might struggle to grasp subtle nuances, sarcasm, or figurative language.
Factual Accuracy
- Models generate responses based on information they learned from their training datasets, but they are not knowledge bases. They may generate incorrect or outdated factual statements.

Ethical Considerations and Risks

The development of vision-language models (VLMs) raises several ethical concerns. In creating an open model, we have carefully considered the following:

Bias and Fairness
- VLMs trained on large-scale, real-world text and image data can reflect socio-cultural biases embedded in the training material. Gemma 4 models underwent careful scrutiny, input data pre-processing, and post-training evaluations as reported in this card to help mitigate the risk of these biases.
Misinformation and Misuse
- VLMs can be misused to generate text that is false, misleading, or harmful.
- Guidelines are provided for responsible use with the model, see the Responsible Generative AI Toolkit.
Transparency and Accountability
- This model card summarizes details on the models' architecture, capabilities, limitations, and evaluation processes.
- A responsibly developed open model offers the opportunity to share innovation by making VLM technology accessible to developers and researchers across the AI ecosystem.

Risks identified and mitigations:

Generation of harmful content: Mechanisms and guidelines for content safety are essential. Developers are encouraged to exercise caution and implement appropriate content safety safeguards based on their specific product policies and application use cases.
Misuse for malicious purposes: Technical limitations and developer and end-user education can help mitigate against malicious applications of VLMs. Educational resources and reporting mechanisms for users to flag misuse are provided.
Privacy violations: Models were trained on data filtered for removal of certain personal information and other sensitive data. Developers are encouraged to adhere to privacy regulations with privacy-preserving techniques.
Perpetuation of biases: It's encouraged to perform continuous monitoring (using evaluation metrics, human review) and the exploration of de-biasing techniques during model training, fine-tuning, and other use cases.

Benefits

At the time of release, this family of models provides high-performance open vision-language model implementations designed from the ground up for responsible AI development compared to similarly sized models.

Citation

If you find our work helpful, please consider citing it:

bibtex
@misc{gemmateam2026gemma4,
      title={Gemma 4 Technical Report}, 
      author={Gemma Team},
      year={2026},
      eprint={2607.02770},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2607.02770}, 
}

Load it the same way

python
from transformers import AutoProcessor, Gemma4UnifiedForConditionalGeneration

repo_id = "wnfldchen/gemma-4-12B-it-qat-w4a16-ct-heretic"

processor = AutoProcessor.from_pretrained(repo_id)
model = Gemma4UnifiedForConditionalGeneration.from_pretrained(
    repo_id,
    device_map="auto",
)

What is included

Compressed model weights in standard Hugging Face transformers layout.
config.json, tokenizer files, processor_config.json, and chat_template.jinja.
The exact compression recipe as recipe.yaml.

Provenance

Source uncompressed Heretic model: wnfldchen/gemma-4-12B-it-qat-q4_0-unquantized-heretic
Quantization: 4-bit int grouped weights, W4A16-style compressed tensors
Group size: 32
Saved format: pack-quantized
Transformers model type: gemma4_unified
Compression recipe preserved as recipe.yaml

W4A16 reproduction

[!TIP] This model is reproducible.

Recreate the public unquantized Heretic source checkpoint pinned below, run the recorded compression command, and verify the resulting artifacts against the hashes in this section.

Source checkpoint

Source unquantized Heretic model: wnfldchen/gemma-4-12B-it-qat-q4_0-unquantized-heretic
Source repo revision pinned at upload time: ca3b6a31f6566df945912929a5964e6d91e918c9
Source model card: README
Source reproducibility guide: README
Compression recipe preserved in this repo as recipe.yaml

Compression command

This is the command to recreate this compressed export from the unquantized Heretic source model:

bash
python scripts/compress_to_w4a16.py \
  wnfldchen/gemma-4-12B-it-qat-q4_0-unquantized-heretic \
  --output ./gemma-4-12B-it-qat-w4a16-ct-heretic \
  --device-map cuda:0

The helper defaults --dtype to auto; the recipe itself is preserved as recipe.yaml.

Compression environment

Platform: Linux-6.8.0-111-generic-x86_64-with-glibc2.39

Table with columns: Component, Version
Component	Version
`Python`	`3.12.13`
`torch`	`2.12.0+cu130`
`transformers`	`5.10.1`
`llmcompressor`	`0.12.0`
`compressed-tensors`

Verification

Recreate the export, then compare the generated files with the SHA256 hashes below.

README.md is intentionally omitted because this model card contains the reproducibility data and therefore cannot hash itself deterministically.

Table with columns: File, SHA256
File	SHA256
`chat_template.jinja`	`ae53464bf3be25802b3a5b37def7fd89667067d7577049b3b2d74c4d8de4c6d4`
`config.json`	`90cd1ae2fdab0741f1fbbb47b399e8ec19b6a657c999fd41450dedf1933170ba`
`generation_config.json`	`0f809910f55b332e334eebf13dc415730e316d187c38d1367dc5b4cd5e27fc25`
`model.safetensors`	`261576a4588f5d7d7e0d220f8761cb7bcd71e815f733fc8c3ed3da3d1c2b79cd`

Source model card reference

The section below is copied from the source unquantized Heretic export for reference. That source card already embeds the upstream Google Gemma 4 model card beneath its own note.

Unless it explicitly says otherwise, benchmark tables, hashes, and abliteration details there refer to the pre-compression checkpoint, not this W4A16 repack.

This is a decensored version of google/gemma-4-12B-it-qat-q4_0-unquantized, made using Heretic v1.4.0 with the Arbitrary-Rank Ablation (ARA) method (with row-norm preservation)

[!TIP] This model is reproducible!

See the README in the reproduce directory for more information.

Abliteration parameters

Table with columns: Parameter, Value
Parameter	Value
start_layer_index	19
end_layer_index	42
preserve_good_behavior_weight	0.9174
steer_bad_behavior_weight	0.0004
overcorrect_relative_weight	0.9004
neighbor_count	13

Performance

Table with columns: Metric, This model, Original model (google/gemma-4-12B-it-qat-q4_0-unquantized)
Metric	This model	Original model (google/gemma-4-12B-it-qat-q4_0-unquantized)
Keywords	11/100	99/100
KL divergence	0.0606	0 (by definition)

Evaluation

Delta = edited model minus original model. Negative values mean the edited model scored lower.

Table with columns: Benchmark, Samples, Metric, This model, Original model, Delta, This stderr, Original stderr
Benchmark	Samples	Metric	This model	Original model	Delta	This stderr	Original stderr
WinoGrande	1267	Accuracy	49.49%	50.51%	-1.02 pp	1.41 pp	1.41 pp
PIQA	1838	Accuracy	53.48%

Interpretation

Across these benchmarks, the edited model scores lower on four reported metrics and higher on two relative to the original model.

The largest decline is -1.23 pp on CommonsenseQA accuracy. The largest improvement is +9.94 pp on EQ-Bench parseable outputs.

Notes

WinoGrande, PIQA, and CommonsenseQA are reported as accuracy-style metrics.
PIQA includes both standard accuracy and normalized accuracy, following lm_eval output.
EQ-Bench is reported with both the benchmark's main score and the percentage of parseable outputs.
Percentage-style standard errors are shown in percentage points.
These results reflect this specific benchmark run and evaluation configuration; broader capability conclusions should ideally be supported with additional benchmarks.

[!NOTE] The remainder of this README reproduces the upstream model card for the base checkpoint. Any benchmark tables or capability claims below this point refer to the upstream base model, not this Heretic-edited derivative, unless explicitly stated.

[!Note] This model card is for the new versions of the Gemma 4 family optimized with Quantization-Aware Training (QAT), which allows preserving similar quality to bfloat16 while dramatically reducing the memory requirements to load the model. Four versions of the QAT checkpoints are available:

Unquantized QAT checkpoints (Q4_0): Half-precision weights extracted from the QAT pipeline, ideal for custom downstream compilation and research. Available for Gemma 4 E2B, E4B, 12B, 26B A4B, and 31B, and their drafter models.

GGUF (Q4_0): Ready-to-deploy formats for broad ecosystem compatibility. Available for Gemma 4 E2B, E4B, 12B, 26B A4B, and 31B.

Mobile-optimized (wNa8o8): A custom schema engineered explicitly for mobile hardware efficiency. It features targeted 2-bit decoding layers, optimized KV caches, and static activations to maximize VRAM savings. Available for Gemma 4 E2B and E4B.

Compressed Tensors (w4a16): QAT checkpoints serialized in the compressed-tensors format for native, optimized inference with vLLM. Available for Gemma 4 E2B, E4B, 12B, and 31B.

Assistant Compatibility: When using multi-token prediction (speculative decoding) with an assistant model alongside a QAT target model, the assistant model must also be a QAT checkpoint with the same precision to ensure compatibility.

Gemma 4 introduces key capability and architectural advancements:

Reasoning – All models in the family are designed as highly capable reasoners, with configurable thinking modes.
Extended Multimodalities – Processes Text, Image with variable aspect ratio and resolution support (all models), Video, and Audio (featured natively on the E2B, E4B, and 12B models).
Diverse & Efficient Architectures – Offers Dense and Mixture-of-Experts (MoE) variants of different sizes for scalable deployment.
Optimized for On-Device – Smaller models are specifically designed for efficient local execution on laptops and mobile devices.
Increased Context Window – The small models feature a 128K context window, while the medium models support 256K.
Enhanced Coding & Agentic Capabilities – Achieves notable improvements in coding benchmarks alongside native function-calling support, powering highly capable autonomous agents.
Native System Prompt Support – Gemma 4 introduces native support for the system role, enabling more structured and controllable conversations.

Models Overview

Dense Models

Table with columns: Property, E2B, E4B, 12B Unified, 31B Dense
Property	E2B	E4B	12B Unified	31B Dense
Total Parameters	2.3B effective (5.1B with embeddings)	4.5B effective (8B with embeddings)	11.95B	30.7B
Layers	35	42	48	60
Sliding Window	512 tokens	512 tokens	1024 tokens	1024 tokens

Mixture-of-Experts (MoE) Model

Table with columns: Property, 26B A4B MoE
Property	26B A4B MoE
Total Parameters	25.2B
Active Parameters	3.8B
Layers	30
Sliding Window	1024 tokens
Context Length	256K tokens
Vocabulary Size	262K
Expert Count	8 active / 128 total and 1 shared
Supported Modalities	Text, Image

Benchmark Results

Table with columns: Gemma 4 31B, Gemma 4 26B A4B, Gemma 4 12B Unified, Gemma 4 E4B, Gemma 4 E2B, Gemma 3 27B (no think)
	Gemma 4 31B	Gemma 4 26B A4B	Gemma 4 12B Unified	Gemma 4 E4B	Gemma 4 E2B	Gemma 3 27B (no think)
MMLU Pro	85.2%	82.6%	77.2%	69.4%	60.0%	67.6%
AIME 2026 no tools	89.2%	88.3%	77.5%	42.5%	37.5%

*Excluding Chinese language.

Core Capabilities

Gemma 4 models handle a broad range of tasks across text, vision, and audio. Key capabilities include:

Thinking – Built-in reasoning mode that lets the model think step-by-step before answering.
Long Context – Context windows of up to 128K tokens (E2B/E4B) and 256K tokens (12B, 26B A4B/31B).
Image Understanding – Object detection, Document/PDF parsing, screen and UI understanding, chart comprehension, OCR (including multilingual), handwriting recognition, and pointing. Images can be processed at variable aspect ratios and resolutions.
Video Understanding – Analyze video by processing sequences of frames.
Interleaved Multimodal Input – Freely mix text and images in any order within a single prompt.
Function Calling – Native support for structured tool use, enabling agentic workflows.
Coding – Code generation, completion, and correction.
Multilingual – Out-of-the-box support for 35+ languages, pre-trained on 140+ languages.
Audio (E2B, E4B, and 12B only) – Automatic speech recognition (ASR) and speech-to-translated-text translation across multiple languages.

Getting Started

You can use all Gemma 4 models with the latest version of Transformers. To get started, install the necessary dependencies in your environment:

pip install -U transformers torch accelerate

Once you have everything installed, you can proceed to load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID,
    dtype="auto",
    device_map="auto"
)

Once the model is loaded, you can start generating output:

python
# Prompt
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Write a short joke about saving RAM."},
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
    enable_thinking=False
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=1024)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

To enable reasoning, set enable_thinking=True and the parse_response function will take care of parsing the thinking output.

Below, you will also find snippets for processing audio (E2B, E4B, 12B only), images, and video alongside text:

Make sure to install the following packages:

pip install -U transformers torch torchvision librosa accelerate

You can then load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, 
    dtype="auto", 
    device_map="auto"
)

Once the model is loaded, you can start generating output by directly referencing the audio URL in the prompt:

python
# Prompt - add audio after text
messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Transcribe the following speech segment in its original language. Follow these specific instructions for formatting the answer:\n* Only output the transcription, with no newlines.\n* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three."},
            {"type": "audio", "audio": "https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/journal1.wav"},
        ]
    }
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=512)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

Make sure to install the following packages:

pip install -U transformers torch torchvision accelerate

You can then load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, 
    dtype="auto", 
    device_map="auto"
)

Once the model is loaded, you can start generating output by directly referencing the image URL in the prompt:

python
# Prompt - add image before text
messages = [
    {
        "role": "user", "content": [
            {"type": "image", "url": "https://raw.githubusercontent.com/google-gemma/cookbook/refs/heads/main/apps/sample-data/GoldenGate.png"},
            {"type": "text", "text": "What is shown in this image?"}
        ]
    }
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=512)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

Make sure to install the following packages:

pip install -U transformers torch torchvision librosa accelerate

You can then load the model with the code below:

python
from transformers import AutoProcessor, AutoModelForMultimodalLM

MODEL_ID = "google/gemma-4-12B-it"

# Load model
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, 
    dtype="auto", 
    device_map="auto"
)

Once the model is loaded, you can start generating output by directly referencing the video URL in the prompt:

python
# Prompt - add video before text
messages = [
    {
        'role': 'user',
        'content': [
            {"type": "video", "video": "https://github.com/bebechien/gemma/raw/refs/heads/main/videos/ForBiggerBlazes.mp4"},
            {'type': 'text', 'text': 'Describe this video.'}
        ]
    }
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=512)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response)

Best Practices

For the best performance, use these configurations and best practices:

1. Sampling Parameters

Use the following standardized sampling configuration across all use cases:

temperature=1.0
top_p=0.95
top_k=64

2. Thinking Mode Configuration

Compared to Gemma 3, the models use standard system, assistant, and user roles. To properly manage the thinking process, use the following control tokens:

Trigger Thinking: Thinking is enabled by including the <|think|> token at the start of the system prompt. To disable thinking, remove the token.
Standard Generation: When thinking is enabled, the model will output its internal reasoning followed by the final answer using this structure:
<|channel>thought\n[Internal reasoning]<channel|>
Disabled Thinking Behavior: For all models except for the E2B and E4B variants, if thinking is disabled, the model will still generate the tags but with an empty thought block:
<|channel>thought\n<channel|>[Final answer]

[!Note] Note that many libraries like Transformers and llama.cpp handle the complexities of the chat template for you.

3. Multi-Turn Conversations

No Thinking Content in History: In multi-turn conversations, the historical model output should only include the final response. Thoughts from previous model turns must not be added before the next user turn begins, with the exception of tool call turns where thinking content should be preserved.

4. Modality order

For optimal performance with multimodal inputs, place:

Image content before the text in your prompt.
Audio content after the text in your prompt.

5. Variable Image Resolution

The supported token budgets are: 70, 140, 280, 560, and 1120.
- Use lower budgets for classification, captioning, or video understanding, where faster inference and processing many frames outweigh fine-grained detail.
- Use higher budgets for tasks like OCR, document parsing, or reading small text.

6. Audio

Use the following prompt structures for audio processing:

Audio Speech Recognition (ASR)

text
Transcribe the following speech segment in {LANGUAGE} into {LANGUAGE} text.

Follow these specific instructions for formatting the answer:
* Only output the transcription, with no newlines.
* When transcribing numbers, write the digits, i.e. write 1.7 and not one point seven, and write 3 instead of three.

Automatic Speech Translation (AST)

text
Transcribe the following speech segment in {SOURCE_LANGUAGE}, then translate it into {TARGET_LANGUAGE}.
When formatting the answer, first output the transcription in {SOURCE_LANGUAGE}, then one newline, then output the string '{TARGET_LANGUAGE}: ', then the translation in {TARGET_LANGUAGE}.

7. Audio and Video Length

Model Data

Data used for model training and how the data was processed.

Training Dataset

Web Documents: A diverse collection of web text ensures the model is exposed to a broad range of linguistic styles, topics, and vocabulary. The training dataset includes content in over 140 languages.
Code: Exposing the model to code helps it to learn the syntax and patterns of programming languages, which improves its ability to generate code and understand code-related questions.
Mathematics: Training on mathematical text helps the model learn logical reasoning, symbolic representation, and to address mathematical queries.
Images: A wide range of images enables the model to perform image analysis and visual data extraction tasks.

The combination of these diverse data sources is crucial for training a powerful multimodal model that can handle a wide variety of different tasks and data formats.

Data Preprocessing

Here are the key data cleaning and filtering methods applied to the training data:

CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was applied at multiple stages in the data preparation process to ensure the exclusion of harmful and illegal content.
Sensitive Data Filtering: As part of making Gemma pre-trained models safe and reliable, automated techniques were used to filter out certain personal information and other sensitive data from training sets.
Additional methods: Filtering based on content quality and safety in line with our policies.

Ethics and Safety

Evaluation Approach

Content related to child sexual abuse material and exploitation
Dangerous content (e.g., promoting suicide, or instructing in activities that could cause real-world harm)
Sexually explicit content
Hate speech (e.g., dehumanizing members of protected groups)
Harassment (e.g., encouraging violence against people)

Evaluation Results

Usage and Limitations

These models have certain limitations that users should be aware of.

Intended Usage

Content Creation and Communication
- Text Generation: These models can be used to generate creative text formats such as poems, scripts, code, marketing copy, and email drafts.
- Chatbots and Conversational AI: Power conversational interfaces for customer service, virtual assistants, or interactive applications.
- Text Summarization: Generate concise summaries of a text corpus, research papers, or reports.
- Image Data Extraction: These models can be used to extract, interpret, and summarize visual data for text communications.
- Audio Processing and Interaction: The E2B, E4B, and 12B models can analyze and interpret audio inputs, enabling voice-driven interactions and transcriptions.
Research and Education
- Natural Language Processing (NLP) and VLM Research: These models can serve as a foundation for researchers to experiment with VLM and NLP techniques, develop algorithms, and contribute to the advancement of the field.
- Language Learning Tools: Support interactive language learning experiences, aiding in grammar correction or providing writing practice.
- : Assist researchers in exploring large bodies of text by generating summaries or answering questions about specific topics.

Limitations

Training Data
- The quality and diversity of the training data significantly influence the model's capabilities. Biases or gaps in the training data can lead to limitations in the model's responses.
- The scope of the training dataset determines the subject areas the model can handle effectively.
Context and Task Complexity
- Models perform well on tasks that can be framed with clear prompts and instructions. Open-ended or highly complex tasks might be challenging.
- A model's performance can be influenced by the amount of context provided (longer context generally leads to better outputs, up to a certain point).
Language Ambiguity and Nuance
- Natural language is inherently complex. Models might struggle to grasp subtle nuances, sarcasm, or figurative language.
Factual Accuracy
- Models generate responses based on information they learned from their training datasets, but they are not knowledge bases. They may generate incorrect or outdated factual statements.

Ethical Considerations and Risks

The development of vision-language models (VLMs) raises several ethical concerns. In creating an open model, we have carefully considered the following:

Bias and Fairness
- VLMs trained on large-scale, real-world text and image data can reflect socio-cultural biases embedded in the training material. Gemma 4 models underwent careful scrutiny, input data pre-processing, and post-training evaluations as reported in this card to help mitigate the risk of these biases.
Misinformation and Misuse
- VLMs can be misused to generate text that is false, misleading, or harmful.
- Guidelines are provided for responsible use with the model, see the Responsible Generative AI Toolkit.
Transparency and Accountability
- This model card summarizes details on the models' architecture, capabilities, limitations, and evaluation processes.
- A responsibly developed open model offers the opportunity to share innovation by making VLM technology accessible to developers and researchers across the AI ecosystem.

Risks identified and mitigations:

Generation of harmful content: Mechanisms and guidelines for content safety are essential. Developers are encouraged to exercise caution and implement appropriate content safety safeguards based on their specific product policies and application use cases.
Misuse for malicious purposes: Technical limitations and developer and end-user education can help mitigate against malicious applications of VLMs. Educational resources and reporting mechanisms for users to flag misuse are provided.
Privacy violations: Models were trained on data filtered for removal of certain personal information and other sensitive data. Developers are encouraged to adhere to privacy regulations with privacy-preserving techniques.
Perpetuation of biases: It's encouraged to perform continuous monitoring (using evaluation metrics, human review) and the exploration of de-biasing techniques during model training, fine-tuning, and other use cases.

Benefits

Citation

If you find our work helpful, please consider citing it:

bibtex
@misc{gemmateam2026gemma4,
      title={Gemma 4 Technical Report}, 
      author={Gemma Team},
      year={2026},
      eprint={2607.02770},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2607.02770}, 
}

gemma-4-12B-it-qat-w4a16-ct-heretic

README

Load it the same way

What is included

Provenance

W4A16 reproduction

Source checkpoint

Compression command

Compression environment

Verification

Source model card reference

This is a decensored version of google/gemma-4-12B-it-qat-q4_0-unquantized, made using Heretic v1.4.0 with the Arbitrary-Rank Ablation (ARA) method (with row-norm preservation)

Abliteration parameters

Performance

Evaluation

Interpretation

Notes

Models Overview

Dense Models

Mixture-of-Experts (MoE) Model

Benchmark Results

Core Capabilities

Getting Started

Best Practices

1. Sampling Parameters

2. Thinking Mode Configuration

3. Multi-Turn Conversations

4. Modality order

5. Variable Image Resolution

6. Audio

7. Audio and Video Length

Model Data

Training Dataset

Data Preprocessing

Ethics and Safety

Evaluation Approach

Evaluation Results

Usage and Limitations

Intended Usage

Limitations

Ethical Considerations and Risks

Benefits

Citation

Explore FriendliAI today

README

Load it the same way

What is included

Provenance

W4A16 reproduction

Source checkpoint

Compression command

Compression environment

Verification

Source model card reference

This is a decensored version of google/gemma-4-12B-it-qat-q4_0-unquantized, made using Heretic v1.4.0 with the Arbitrary-Rank Ablation (ARA) method (with row-norm preservation)

Abliteration parameters

Performance

Evaluation

Interpretation

Notes

Models Overview

Dense Models

Mixture-of-Experts (MoE) Model

Benchmark Results

Core Capabilities

Getting Started

Best Practices

1. Sampling Parameters

2. Thinking Mode Configuration

3. Multi-Turn Conversations

4. Modality order

5. Variable Image Resolution

6. Audio

7. Audio and Video Length

Model Data

Training Dataset

Data Preprocessing

Ethics and Safety

Evaluation Approach

Evaluation Results